My general feeling on the review article that Kristof references is that such a review, as the authors state, was strongly needed.
My worries, based on work I did ten years ago to make quantitative connections between toxicological results on many drugs and human health, are that the experiments have larger error bars than the authors claim that they do, that the different experimental studies had different methods and thus may not be comparable, and that non-monotonic response may be a politically correct way to get more government funding but not actually supported very well by the results.
Whether it is EDCs or other chemicals that modulate hormonal, toxic response, or DNA damage, low level dose studies seem to be very hard to do well. There are too many uncontrolled variables including genetic background, life experiences, other diseases, current health, humidity, and other confounding variables in the sample population. There is also the problem, especially in toxicology studies, that results in one species are often not predictive of results in other species, often for reasons that are neither clear nor predictable. I wrote a 300 page book on trying to connect microarray results to human toxicity of novel drugs. The main conclusion of this research was that the mathematics to connect the microarray results with toxicology results was complex and, at the time, the actual results from both microarrays and toxicology were not reproducible enough that it was worth the effort to try to connect them through some statistically robust function.
Low levels of EDCs seem to be correlated with misdevelopment in boys and girls. The girls go through puberty early while the boys may become feminized. The effect starts as early as when the children are in the womb. My current source for this is pages 99-115 of Leonard Sax's book "boys adrift" and references therein.
This story has gotten more complex with the finding that endocrine disrupters broken down by sunlight can reform in the dark.
Here is a list of the worst twelve EDCs.
First posted on 12 May 2012.